Phonon: small open voice model that speaks on the device, offline

Gradium has released Phonon, text-to-speech models of about 65 MB and 75 MB that run on phones, laptops, single-board computers and in a browser tab with no server or network. The English-only Phonon 31m is Apache 2.0, so it is free for commercial use; Phonon 40m adds French, German, Spanish and Portuguese under a non-commercial licence, with commercial pricing based on number of users. Gradium says it ranks in the top three for intelligibility in every language it evaluated on Hugging Face's Open TTS Leaderboard, first in German and Spanish, and tops the CPU streaming test on time to first audio. It has more than 100 built-in voices, is in Transformers from v5.21, and was downloaded more than 200,000 times in its first two weeks.